Urs raised the legitimate point that it looks like that our partial reconstruction for D*lnu is 100% efficient: we claim we kill 68% of the events which is exactly the fraction of D*lnu events where a soft pion exists. I reobtained the result in a slightly different way (and actually got 70%, but with some error...) http://www.slac.stanford.edu/~rfaccini/phys/vub/deltam/dmfraction.eps shows the distribution of prmm2 for the D*lnu events (as flagged at MC truth level) after all cuts. I just counted the fraction of events above -3. I think that this effect can be explained because I already require the soft pion to be present (qtot=0) and at this point the resolution is good enough to contain all events. In addition to this the sample might be a bit biased against D*->Dpi0 decays, due to the mm2 cut. finally some of the killed events belong to the combinatorics, thus compensating for possible inefficiencies. ciao ric